AI Paper Atlas

Research map / Learning theory and algorithms / Kernels and regression theory

Rethinking Generalization of Neural Models: A Named Entity Recognition Case Study

Jinlan Fu, Pengfei Liu, Qi Zhang

AAAI 2020 · Learning theory and algorithms › Kernels and regression theory · approach: Empirical generalization measures · 74 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Correct human annotation errors identified using entity coverage ratioNER models evaluated on CoNLL · ReCoNLLF1: 93.03 -> 93.78Improved
Combine character-level FLAIR with word-level GloVeLSTM-CRF NER · CoNLLF1: 91.79 -> 93.03Improved
Use character-level pretrained embeddings rather than randomly initialized word embeddingsNER models · WNUTF1: 14.97 -> 41.49Improved
Use FLAIR + GloVe rather than randomly initialized word embeddingsNER models · WNUTF1: 14.97 -> 45.96Improved
Use ELMo or FLAIR rather than simpler embedding choicesNER models · CoNLL and WNUT, seen entitiesF1No clear effect
Train with entities seen under the same label rather than entities with no same-label training occurrenceNER models · CoNLL and WNUTF1: 98.56 -> 73.58Improved
Select source domains using entity and contextual coverage rather than adding training data indiscriminatelyCross-domain NER · WNUTF1Improved
Increase the amount of training data without prioritizing source-domain relevanceCross-domain NER · WNUTF1Mixed
Use contextual coverage in addition to entity coverage when choosing training domainsCross-domain NER · OntoNotes-wbF1: 13.8 -> 21.8Improved
Use a different architecture or pretrained knowledge sourceNER models · Hard cross-category caseserror ratioNo clear effect
Train on examples from categories with conflicting gradient directionsNER models · Overlapping entity categoriescross-category error ratioHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy